Original instruction-following text model, distinct from the newer Command R and Command A families. Cohere deprecated the original model in September 2025; Oracle serving modes have separate retirement dates.
Bedrock temperature 0–5, default 0.9, Bedrock top-p 0–1, default 0.75; top-k 0–500, default 0, Bedrock max_tokens 1–4096, default 20, 1–5 generations, default 1; up to four stop sequences, Truncation NONE/START/END; END is default, Token likelihoods GENERATION/ALL/NONE and logit bias −10 to 10Source for Generation controls (opens in a new tab)
These are the Bedrock contract, not universal Cohere defaults. The direct legacy Generate API lists temperature default 0.75 and max_tokens unset by default.
Customization options
Original Command models supported fine-tuning on OCI. Cohere retired its own dashboard/API fine-tuning and access to affected existing fine-tuned models in September 2025.Source for Customization options (opens in a new tab)
OCI’s historical customization support is separately documented in its service GA announcement. Private contracts and host lifecycles are distinct.
Access & pricing
Prices belong to an endpoint and billing tier. A provider listing does not guarantee access for every account or region.
Prices retain their published currencies and billing units below.
Published billing units · original currency · observed 14 Oct 2026
Category
Usage
Rate
Unit
Historical pricing
Input tokens1
$0.30
1M tokens
Historical pricing
Output tokens1
$0.60
1M tokens
Historical pricing
Training tokens1
$1.00
1M training tokens
Historical pricing
Custom model storage1
$1.95
model-month
Historical pricing
Custom model inference without commitment1
$8.56
model-unit hour
Rate conditions
US East (N. Virginia); AWS price-map publication 2026-10-12T08:40:08Z. Retained original-Command meter. No active availability or per-question quote is inferred; region, customization access and billing terms must be verified independently.
On-demand retired 2024-10-02; new dedicated base-model cluster creation retired 2025-08-07. Oracle says existing active dedicated clusters can continue, and custom models derived from retired bases have separate continuation rules.
Dedicated date independently documented in the linked dedicated retirement table.
Endpoint pricing has not been collected. Check the provider for applicable rates.
Data & enterprise terms
Retention, training use, regional processing and enterprise terms have not been summarized for this endpoint.
Caching
Caching can reuse earlier input processing. Availability, lifetime and billing depend on the endpoint. A cache hit still generates a new response.
01 / INITIAL INPUTProcess & retain
The system processes eligible input and retains reusable state.
02 / LATER REQUESTCheck available state
Reuse depends on matching input, session state and the cache policy.
03 / CACHE HITReuse & generate
Reusing state can reduce repeated work. Any charges follow the billing rules below.
Cohere API (deprecated)
command-light
Not verified
No original-model prompt-cache contract verified. Newer Command/Embed capabilities are not inherited.
How it works
Not collected
Minimum reusable input
Not collected
Lifetime & refresh
Not collected
How to enable it
Not collected
Usage fields
Not collected
Billing rules
Not collected
Limits & data boundaries
Not collected
No verified cache rate for this endpoint. An empty rate is not a zero charge.
Caching reviewed 14 Oct 2026
Cohere Generate API (historical alias)
command-medium
Not verified
No original-model prompt-cache contract verified. Newer Command/Embed capabilities are not inherited.
How it works
Not collected
Minimum reusable input
Not collected
Lifetime & refresh
Not collected
How to enable it
Not collected
Usage fields
Not collected
Billing rules
Not collected
Limits & data boundaries
Not collected
No verified cache rate for this endpoint. An empty rate is not a zero charge.
Caching reviewed 14 Oct 2026
Cohere API (historical experimental channel)
command-light-nightly
Not verified
No original-model prompt-cache contract verified. Newer Command/Embed capabilities are not inherited.
How it works
Not collected
Minimum reusable input
Not collected
Lifetime & refresh
Not collected
How to enable it
Not collected
Usage fields
Not collected
Billing rules
Not collected
Limits & data boundaries
Not collected
No verified cache rate for this endpoint. An empty rate is not a zero charge.
Caching reviewed 14 Oct 2026
Amazon Bedrock (historical model)
cohere.command-light-text-v14
Not verified
No original-model prompt-cache contract verified. Newer Command/Embed capabilities are not inherited.
How it works
Not collected
Minimum reusable input
Not collected
Lifetime & refresh
Not collected
How to enable it
Not collected
Usage fields
Not collected
Billing rules
Not collected
Limits & data boundaries
Not collected
No verified cache rate for this endpoint. An empty rate is not a zero charge.
Caching reviewed 14 Oct 2026
Oracle OCI Generative AI (retired base model)
cohere.command-light
Not verified
No original-model prompt-cache contract verified. Newer Command/Embed capabilities are not inherited.
How it works
Not collected
Minimum reusable input
Not collected
Lifetime & refresh
Not collected
How to enable it
Not collected
Usage fields
Not collected
Billing rules
Not collected
Limits & data boundaries
Not collected
No verified cache rate for this endpoint. An empty rate is not a zero charge.
Caching reviewed 14 Oct 2026
What the measurements say
No benchmark observations have been added for this model. We do not turn a vendor description into a performance score.
Latency and throughput depend on the host, request size, reasoning settings and load. There is no measured speed or global usage figure in this record.
The record behind the model
Why it is included
Historical distribution inference: Cohere identifies this original model and its predecessor name as an API model in the official rename notice. The archived product documentation exposes managed inference, not an original weight release. Reviewed Cohere/CohereLabs repositories do not provide these original weights; this is not a blanket license assertion for newer public Command R/A families.
